dissenting:
In this case we return, unhappily, to rent control. I respectfully but emphatically dissent.
We took a wrong turn in Richardson v. City and County of Honolulu, 124 F.3d 1150, 1164 (9th Cir.1997), where we held that when a tenant actually captures a premium resulting from a rent control statute, the statute is an unconstitutional regulatory taking unless it “substantially furthers a legitimate state interest.” We continued on the wrong path in Chevron USA, Inc. v. Cayetano, 224 F.3d 1030 (9th Cir.2000) (Chevron I), where we concluded that when a tenant may possibly capture a premium, the “substantially furthers” test must be satisfied and remanded for trial. Id. at 1040, 1042 (now stating the test as whether the statute “substantially advances a legitimate state interest”) (emphasis added). We persisted on that wrong path in Chevron USA Inc. v. Lingle, 363 F.3d 846 (9th Cir.2003) (Chevron II), where we affirmed the judgment of the district court, concluding that its factual finding that the rent control statute did not satisfy the “substantially advances” test was not clearly erroneous. 363 F.3d at 857. We now desert the wrong path taken in Richardson, Chevron I, and Chevron II, only to take yet another wrong path.
I disagreed with the panel in both Chevron cases. See Chevron I, 224 F.3d at 1042 (W. Fletcher, J., concurring in the judgment); Chevron II, 363 F.3d at 858 (W. Fletcher, J., dissenting). But if the majority in today’s case were faithful to those cases, I would concur because they are (or were) the law of the circuit.
However, the majority today refuses to apply the Chevron cases. Under those cases, the possibility of a premium capture triggered the application of the “substantially advances” test. But under today’s holding, what was the trigger for the “substantially advances” test has itself become the test. Under today’s holding, if there is a possibility that tenants will capture a premium, or even part of a premium, that possibility in and of itself renders a rent control ordinance unconstitutional.
I. Procedural Background
The City of Cotati, located in Sonoma County, California, north of San Francisco Bay, has controlled rents in mobile home parks since 1979. In 1998, Cotati adopted Ordinance 680, amending the previous rent control ordinance but continuing the same general scheme. Plaintiffs-Appellants Cashman and Sutsos (“plaintiffs”) own mobile home parks in Cotati. In 1999, they brought a facial challenge to Ordinance 680, contending that it effects a regulatory taking in violation of the Due Process Clause of the Fourteenth Amendment.
In 2000, based on its reading of Richardson, the district court granted summary judgment to plaintiffs. It held that in the absence of a mechanism in Ordinance 680 preventing mobile home park tenants from capturing a premium, the entire ordinance effected an unconstitutional taking. The district court subsequently amended its judgment under Federal Rule of Civil Procedure 59(c), holding unconstitutional only the vacancy control portion of Ordinance 680. After the entry of the district court’s amended judgment, we decided Chevron I, in which we remanded to the district court for a determination at trial whether a rent control statute “substantially advance[d] a legitimate state interest.” Chevron I, 224 F.3d at 1040, 1042.
After Chevron I was handed down, the district court vacated its summary judgment in this case. Doing as we had in*901structed the district court to do in Chevron I, the district court held a trial to determine whether Ordinance 680 satisfies the “substantially advances” test. After a five-day trial, during which evidence was presented by experts from both sides, the district court held against the plaintiffs.
First, the district court held that plaintiffs’ claims were barred by the statute of limitations because the 1979 ordinance was applied in the same manner as the later-enacted Ordinance 680. Thus, the one-year statute of limitations had begun to run in 1979. Second, the court held that if the statute of limitations had not run, plaintiffs’ claims failed on the merits. It held that plaintiffs had not shown by a preponderance of the evidence that Ordinance 680 is likely to create a premium. It held, further, that even if a premium did exist, plaintiffs had not shown by a preponderance of the evidence that such a premium would prevent the ordinance from substantially advancing such legitimate state purposes as protecting tenants, including elderly and low-income people, from excessive rent increases.
On appeal, the majority first holds that the statute of limitations did not begin to run until the passage of Ordinance 680 in 1998. Although it is a close question, I do not quarrel with the majority on this point.
The majority next holds that the district court erred in amending and then vacating its summary judgment. The majority concludes that there is “the possibility of a premium” under Ordinance 680. Maj. op. at 899. In the sense used here, a “premium” is capitalized value conferred on a tenant by virtue of rent control. If rent control lowers the price of a tenancy below the open-market price for a comparable tenancy, and if an existing tenant can sublease or otherwise sell a rent-controlled tenancy, the existing tenant may be able to obtain, as part of the selling price for that tenancy, the capitalized value of the difference between the rent-controlled rent and the open-market rent. This phenomenon is known as “capturing the premium.” Depending on the circumstances and the nature of the particular rent control regime, a tenant may be able to capture some, or even all, of the premium resulting from rent control.
When the majority writes that there is “the possibility of a premium,” its use of the indefinite article “a” is important. “The premium” would mean the entire premium resulting from rent control. By contrast, “a premium,” as used by the majority, means any part of the premium. The majority holds that because of the possibility that mobile home owners might capture “a premium” — that is, part of the premium resulting from Ordinance 680 — plaintiffs “are entitled to summary judgment on their facial claim unless the City presents sufficient evidence of external factors that will prevent the existence of a premium altogether.” Id. at 899. In other words, the majority requires that, in order to sustain the constitutionality of Ordinance 680, defendant Cotati must prove that mobile home owners would not capture any part of a premium as a result of the ordinance. The majority concludes that there is no genuine issue of material fact as to whether mobile home owners would capture any part of such a premium, and therefore reinstates the district court’s pre-Chevron I summary judgment that Ordinance 680 is unconstitutional.
II. Rent Control in Mobile Home Parks
Residential rentals in mobile home parks are different from ordinary residential rentals. In an ordinary residential rental, the landlord owns the apartment or house, and rents that apartment or house to the tenant. The tenant brings to the house or apartment only his or her own *902personal effects and, at the end of the tenancy, takes away those personal effects. If a landlord raises the rent on the apartment or house and the tenant moves out to avoid the higher rent, the tenant incurs only the expense of moving those effects to a different apartment or house.
By contrast, in a typical mobile home park rental, the landlord is the “park” owner. The park owner owns only the land, or “pad,” on which the mobile home sits. The tenant rents the pad but owns the mobile home. Despite their name, mobile homes are not, in fact, mobile. They are constructed off site, put on wheels to be transported, and towed to the site where they will be placed. The difficulty and very substantial expense of moving a mobile home once it has been placed on its pad means that it is almost never moved thereafter.
A mobile home owner — that is, a tenant in a mobile home park — is therefore in a very different economic situation from an ordinary residential tenant. Because of the almost prohibitive expense involved in moving a mobile home, a mobile home park owner has considerable freedom, absent rent control, to raise rents on all of his or her existing tenants. To state it in economic terms, if the park owner raises the rent on an existing tenant who wants to continue living in the mobile home, the tenant is likely to pay any rent increase whose capitalized cost is less than the cost of moving the mobile home. Or, if the park owner raises the rent when an existing tenant sells the mobile home to a new owner, the price paid for the mobile home will likely be reduced by the amount of the capitalized value of the rent increase. To state it in colloquial terms, absent rent control, a park owner can gouge existing tenants.
As a result, many jurisdictions in California and elsewhere have adopted rent control ordinances specifically tailored to mobile home parks. Approximately 100 jurisdictions in California have ordinances of this kind. Most such ordinances are like the ordinances Cotati has had since 1979. Cotati has consistently limited rent increases for existing tenants, and, with some exceptions, limited rent increases for purchasers who buy from existing tenants. Ordinance 680, adopted by Cotati in 1998, limits rent increases in frequency to once a year and in amount to the lesser of 6 percent or the change in the Consumer Price Index. Ord. § 19.14.004(a). Under the “vacancy control” provision of the ordinance, when an existing tenant sells his mobile home to a new tenant, the park owner is not permitted to raise the rent. Ord. § 19.14.150(a). However, if a resident is evicted or voluntarily removes his home from the park, the landowner may establish a new base rent for the next tenant. Id.
III. The Majority Errs in Reinstating Plaintiffs’ Summary Judgment
Assuming that our earlier decisions in Richardson, Chevron /, and Chevron II are correct, the majority errs in reinstating the district court’s summary judgment that Ordinance 680 is an unconstitutional regulatory taking. In reinstating that judgment, the majority writes:
The only evidence the City submitted [at summary judgment] is a 1990 report prepared for the City by an expert who stated that mobilehome owners in Sono-ma County (which includes Cotati) are so distrustful of rent control that they will pay less for a mobilehome than for one that is not under rent control. The original district court opinion correctly finds that such evidence is insufficient to create a genuine issue of material fact on a facial takings claim under Richardson and Chevron I. On a facial takings *903claim, a court is to look only to the ordinance’s general scope and dominant features, not its application in specific circumstances.... [T]he mistrust of some would-be mobilehome purchasers in Sonoma County does not pertain to the “dominant” impact of the Ordinance.
Id. at 899. The majority makes several mistakes.
First, the majority misdescribes the nature of a facial challenge. A successful constitutional facial challenge to an ordinance requires that a plaintiff show that the ordinance is incapable of constitutional application under any circumstances. United States v. Salerno, 481 U.S. 739, 745, 107 S.Ct. 2095, 95 L.Ed.2d 697 (1987); Cogswell v. City of Seattle, 347 F.3d 809, 814 (9th Cir.2003). The plaintiff has the burden of proof of unconstitutionality. Id. at 813. Instead of following this established law, the majority concludes that “the possibility” of capture of any part of the premium means that the defendant rather than the plaintiff has the burden of proof of the ordinance’s constitutionality. According to the majority, the defendant’s burden is to show that there are “external factors that will prevent the existence of a premium altogether.”
Second, the majority misapplies Chevron I. In Chevron I, there was a possibility of premium capture by the tenants. We did not, because of that possibility, require the defendant to show external factors that would altogether have prevented the existence of a premium. Nor did we hold that in the absence of such a showing by the defendant the plaintiff was entitled to summary judgment. Rather, we remanded to the district court for trial to determine if the statute “substantially advanced a legitimate state interest.” We wrote that “[bjecause resolution of these factual issues[about the existence of a premium and benefits to tenants] is necessary to determine whether Act 257 substantially advances, or bears a reasonable relationship to, the State’s interest in lowering gasoline prices, the district court erred in granting summary judgment.” 224 F.3d at 1042. Based on Chevron I, the district court in this case was entirely correct to vacate its summary judgment and to hold a trial to resolve the factual issues of whether a premium was created by Ordinance 680, and whether that ordinance substantially advances a legitimate state purpose.
Third, the majority mischaracterizes the evidence before the district court on summary judgment. It states that defendant Cotati put only one “report” into evidence and that this report did not create a genuine issue of material fact. The majority is referring to an extensive 1990 empirical study conducted by Dr. Stephen Lewis, entitled “Economic Impact of Rent Controls on Mobile Home Resale Prices in Sonoma County and Cotati.” Dr. Lewis, who later testified at trial for Cotati, concluded that there was no premium as a result of mobile home rent control in Cota-ti. Indeed, according to Dr. Lewis’s study, mobile homes subject to rent control in Cotati sold at a substantial discount rather than a premium. Dr. Lewis summarized the results of his study as follows:
When mobile home pad rents are subject to rent controls and when these rents are below the market value, economic theory suggests that pad rent differentials will be capitalized into the resale prices of mobile homes subject to rent controls. The data examined in this study leads to the conclusion that rent controls in Sonoma County and more specifically in Cotati and Rohnert Park have not had the expected impact on mobile home resale prices. Using data from several sources and standard statistical techniques, it has been determined after controlling for other factors *904that mobile home parks subject to rent control have sold on average for less than those not subject to controls....
Mobile homes subject to rent controls are less desirable than uncontrolled ones because of uncertainties and risks associated with rent control programs....
[I]t is clear that in Sonoma County mobile homes subject to rent controls do not sell at a premium but rather at a substantial discount. Claims that such premiums exist are unwarranted and without merit.
(Emphasis added.)
Finally, the majority misstates the record in saying that the only evidence presented by Cotati at summary judgment was Dr. Lewis’s study. Cotati presented in addition a study by Dr. Kenneth Baar, who also later testified at trial for Cotati. Dr. Baar studied the effect of vacancy controls in mobile home park rentals in a different community in California. Vacancy controls limit the ability of the park owner to increase pad rent when a mobile home owner sells the mobile home. Such controls have been a feature of Cotati mobile home rent control both before and after the adoption of Ordinance 680. Dr. Baar concluded that a vacancy control feature of a rent control ordinance did not create a premium — that is, it does not produce a higher resale price for mobile home owners. Dr. Baar wrote:
From a purely theoretical perspective, the reduced rents resulting from vacancy control might be precisely offset by the increase in value of the mobilehome in the vacancy controlled space. In fact, however, the markets for the mobile-home space and the mobilehome operate in substantially different fashions, so that effects of vacancy control do not follow the foregoing theoretical model.
From an affordable housing point of view, it appears that in [this community], even with vacancy control, the initial costs of mobilehome ownership for prospective purchasers ... will be lower than these costs in the absence of vacancy control.
(Emphasis added.)
If the majority had properly allocated the burden of proof in plaintiffs’ facial attack on Ordinance 680, if it had properly followed Chevron I, and if it had properly characterized the evidence put in the record by Cotati at summary judgment, it would have concluded that the district court correctly vacated the summary judgment and sent this case to trial. At summary judgment, Dr. Lewis and Dr. Baar presented competent evidence that, if believed, compelled a conclusion that no premium was, or would be, created by mobile home park rent control in Cotati. Because Cotati produced sufficient evidence to create a genuine issue of material fact, the district court was required under Chevron I to deny summary judgment and proceed to trial.
IV. Defendant’s Judgment after Trial Should be Affirmed
In reinstating the district court’s summary judgment, the panel majority implicitly concludes that the evidence at trial was not enough to justify a judgment sustaining the constitutionality of Ordinance 680. I say this because, although Cotati presented more evidence at trial than at summary judgment, that evidence was to the same effect as the Lewis and Baar studies already presented by Cotati at summary judgment. If the majority is willing to disregard the evidence of the Lewis and Baar studies, it would be equally willing to disregard the additional evidence. This additional evidence consists of four empirical studies conducted by James *905Brabant, a professional real estate appraiser, of actual mobile home sales in Sonoma County and adjoining Napa County. Mr. Brabant concluded from these studies that no premium was created by Ordinance 680.
Plaintiffs presented evidence at trial through Dr. Werner Hirsch, Dr. Robert Edelstein, and John Patrick Neet. However, the district court found as a matter of fact that plaintiffs’ experts’ studies and testimony were less believable than those of Dr. Lewis, Dr. Baar, and Mr. Brabant.
I disagree with the majority’s implicit conclusion that the evidence presented by Cotati was insufficient to support the district court’s judgment. I would uphold the factfinding of the district court after trial under the clearly erroneous standard, as we upheld the factfinding of the district court after trial in Chevron II, 363 F.3d at 857, and I would affirm the judgment of the district court.
V. A Return to Judicial Activism
We learned in the 1930s that economic regulation is generally done better by politically accountable legislators than by life-tenured judges. I regret to say that the Ninth Circuit is unlearning that painfully learned lesson.
There are two different constitutional tests that could conceivably apply to a rent control ordinance. The first is the “reasonableness” test ordinarily applied to rent and price control statutes. See, e.g., Pennell v. City of San Jose, 485 U.S. 1, 11, 108 S.Ct. 849, 99 L.Ed.2d 1 (1988) (upholding a rent control ordinance because it was not “arbitrary, discriminatory, or demonstrably irrelevant to the policy the legislature is free to adopt”) (quoting Permian Basin Area Rate Cases, 390 U.S. 747, 769-70, 88 S.Ct. 1344, 20 L.Ed.2d 312 (1968)). The second is the “substantially advances a legitimate state interest” test ordinarily applied to zoning and other land use regulations. See, e.g., Agins v. City of Tiburon, 447 U.S. 255, 261, 100 S.Ct. 2138, 65 L.Ed.2d 106 (1980) (upholding a zoning ordinance because it “substantially advanced] legitimate governmental goals”); Dolan v. City of Tigard, 512 U.S. 374, 385, 114 S.Ct. 2309, 129 L.Ed.2d 304 (1994); Nollan v. California Coastal Comm’n, 483 U.S. 825, 834, 107 S.Ct. 3141, 97 L.Ed.2d 677 (1987).
In Richardson, decided in 1997, we applied the “substantially advances” test to invalidate a rent control ordinance under which we concluded that the tenant was actually able to capture the premium resulting from rent control. See 124 F.3d at 1166 (“Incumbent owner occupants who sell to those who intend to occupy the apartment will charge a premium for the benefit of living in a rent controlled condominium. The price of housing ultimately will remain the same.”) (emphasis added). In Chevron I, three years later, we went beyond Richardson, holding that the “substantially advances” test must be applied to the rent control statute at issue because of the “possibility” of premium capture. See 224 F.3d at 1035 (applying the “substantially advances” test because of the “stipulated possibility that [a tenant] will be able to capture the value of the decreased rent in the form of a premium.”). In Chevron II, just last year, we upheld as not clearly erroneous the factfinding of the district court that the “substantially advances” test had not been satisfied, and that the rent control statute was unconstitutional. See 363 F.3d at 857 (“Based on all the evidence adduced at trial, the district court concluded that [the rent control statute] will not substantially advance a reduction in the retail price of gasoline. The court’s factual findings and conclusions of law are consistent with the views of the parties’ experts and are not clearly erroneous.”). Finally, today the majority holds that the mere “possibility” that a tenant could capture part of the premium *906from a rent control ordinance is enough to render the ordinance unconstitutional. It then reinstates a summary judgment of unconstitutionality despite evidence in the record that no premium actually resulted, or would result, from Ordinance 680.
The only possible basis for the majority’s holding is dictum in Yee v. City of Escondido, 503 U.S. 519, 112 S.Ct. 1522, 118 L.Ed.2d 153 (1992), in which the Supreme Court upheld a mobile home rent control ordinance against a takings challenge. The Court upheld the ordinance against a physical takings challenge, but refused to consider whether the ordinance constituted a regulatory taking. In the course of noting that it was not considering the regulatory takings issue, the Court wrote in dictum:
[T]he effect of the rent control ordinance, coupled with the restrictions on the park owner’s freedom to reject new tenants, is to increase significantly the value of the mobile home. This increased value normally benefits only the tenant in possession at the time the rent control is imposed.... Petitioners are correct in citing the existence of this premium as a difference between the alleged effect of the Escondido ordinance and that of an ordinary apartment rent control statute.... [Petitioners contend that the Escondido ordinance transfers wealth only to the incumbent mobile home owner. This effect might have some bearing on whether the ordinance causes a regulatory taking, as it may shed some light on whether there is a sufficient nexus between the effect of the ordinance and the objectives it is supposed to advance. See Nollan v. California Coastal Comm’n, [483 U.S. 825, 834-35 (1987)]. But it has nothing to do with whether the ordinance causes a physical taking.
Id. at 530, 112 S.Ct. 1522 (emphasis in original).
It is a very long way from this passage in Yee to the majority’s holding in this case. The Court in Yee did not say, even in a case where there was an actual premium, that the Nollan “substantially advances” test would apply. Nor did the Court in Yee say that where there was only the possibility of a premium, the “substantially advances” test would apply. Nor, finally, did the Court in Yee say that the possibility of a premium alone would render a rent control ordinance unconstitutional, without regard to the “substantially advances” test.
Based on a slender hint in Yee, we have constructed a new doctrine out of whole cloth. This interpretation is the Ninth Circuit’s alone; I have found no case outside our circuit that reads Yee as the majority reads it. After today’s decision, the possibility of capture of any part of a premium by a tenant renders a rent control ordinance unconstitutional. Even if two qualified experts present evidence that there is, in actual fact, no premium, a plaintiff is entitled to summary judgment that the ordinance is unconstitutional. With all due respect to my colleagues, this simply cannot be the law.